I need to train a YOLOv8 model to detect objects in some extremely flat and elongated images. But I find that 'rect' mode is incompatible with multiple GPUs, it warns me that 'rect' mode is incompatible with Multi-GPUs, setting rect=False.
I try to set augment=True and add one line code of Resize in the init function of Albumentations class, but it seems like it's wrong. So how can I train a YOLOv8 model with 'rect' mode and multiple GPUs, even if I need to modify the source code of ultralytics? The added code
